Cross-Sectional Momentum
XS-MOMRank the universe. Own the winners. Exit the laggards.
Cross-sectional momentum is one of the most robust and academically documented anomalies in financial markets. Rather than predicting whether any single asset will go up or down, XS-MOM asks a simpler question: which assets are strongest right now, relative to everything else? By continuously scoring the entire universe, ranking the results, and holding only the top performers, the strategy systematically captures relative outperformance — a pattern that persists because markets trend, and trending assets tend to keep trending until they do not.
The Core Concept
At any point in time, we score every Bybit perpetual by its recent price performance relative to the full universe. The top-ranked symbols are our long candidates. This approach exploits relative strength — not absolute direction. A coin can be our top pick even in a flat or slightly declining market, as long as it is declining less than everything else. The ranking resets on every rebalance cycle so stale winners are continuously pruned.
Crypto perpetuals exhibit strong cross-sectional momentum effects, validated in Han, Kang & Ryu (2024). The 24/7 market, continuous price discovery, and retail-dominated flow create persistent momentum windows that systematic strategies can exploit. Funding rates align short interest with momentum, amplifying the signal in strong regimes and providing early warning in reversals.
Lookback Periods
| Strategy | Lookback | Rebalance | Universe |
|---|---|---|---|
| Bot6 | 72 hours | 4h | Top-50 by momentum |
| Bot8 | 14 days | 4h | Top-50 by momentum |
| Bot16 | 5 days | 12h | Top-40 by momentum |
| Bot17 | 21 days | 24h | Top-30 by momentum |
Why run multiple lookbacks simultaneously? MasterBitcoin runs multiple lookback periods in parallel. Longer lookbacks (14–21 days) catch macro-level momentum carried by institutional flows and trend-following funds. Shorter lookbacks (72 hours–5 days) catch tactical entries on fresh breakouts and mean-reversion exits. Combined, they improve signal diversification — a coin that ranks highly across all lookbacks is a high-conviction candidate.
The Scoring Formula
Percentile rank of the asset's N-day return within the active universe, normalized to 0–100. A score of 90 means the asset outperformed 90% of all ranked perpetuals over the lookback window.
Time-series momentum — is the asset trending up on its own? A high TS score means the asset has been consistently advancing on its own price history, independent of how others are doing. Normalized 0–100.
Market structure — price vs VWAP, breadth alignment, and order-book health. Rewards assets trading above their volume- weighted average price with broad market participation behind the move.
Developing Setup
Minimum threshold to appear on the signal board
Strong Signal
High-conviction setup, all three factors aligned
Entry Quality Filters
Raw momentum rank is necessary but not sufficient. Five independent gate layers check market conditions before any entry is allowed. Each gate has a distinct failure mode — together they are orthogonal, meaning they rarely trigger simultaneously and do not compound false positives.
Dispersion Gate
The universe must show a minimum spread between top performers and bottom performers. When everything is moving in unison — either all up or all down — the cross-sectional signal collapses to noise. The dispersion gate ensures the ranking is meaningful: there must be genuine divergence in returns before we act on any rank.
Alt-BTC Spread Gate
Altcoins vs BTC must be in the band -2.5% to -1.5% over 24 hours. When alts are massively underperforming BTC, it signals a flight-to-quality rotation — a hostile environment for altcoin long momentum. When alts are outperforming BTC by a large margin, it can signal a blow-off top. The band identifies the sweet spot: BTC is still outperforming, but alts are not in freefall.
Blow-Off Gate
Blocks entry if the top-ranked asset has moved more than a threshold percentage in the final 1 hour of the lookback window. A coin that gained 7% in the last hour to claim the top rank is not a momentum entry — it is a late-stage extension that is statistically likely to reverse. This gate prevents chasing blow-off tops disguised as momentum signals.
BTC Direction Gate
Blocks all new entries when BTC's 24-hour return is below -2.0%. Grounded in Daniel & Moskowitz (2016) on momentum crashes: momentum strategies suffer maximum drawdown during sharp market reversals. Since BTC drives systemic risk in crypto, a falling BTC environment is the highest-crash-risk state for any long momentum strategy. The gate does not close existing positions — only new entries are blocked.
Rank Skip
Removes the top 2–7 most overextended symbols from the entry pool. Research on Bot8 (14d lookback) showed that ranks 1–7 averaged -$0.35/trade — these symbols had already moved so far that mean-reversion drag outweighed momentum continuation. Ranks 8–15 averaged +$0.74/trade. Note: Bot16 (5d lookback) shows the opposite — ranks 1–5 are the highest-alpha group. Rank skip is calibrated per-lookback, not applied uniformly.
Exit Architecture
Entries are gated aggressively; exits are layered. Five distinct mechanisms can close a position, each targeting a different failure mode. They operate independently — multiple can trigger on the same position.
Native Stop-Loss
Placed directly on Bybit at -25% from entry. Wide to capture the full momentum wave without being stopped out by normal intraday volatility. Bybit executes this as a native order — survives bot restarts, network issues, and any Python-layer failures.
Max Hold Window
Each strategy has a maximum hold window of 72–96 hours depending on the lookback. Positions held beyond this window are closed regardless of P&L. Prevents zombie positions that survived but are no longer driven by the original momentum signal.
Skip-Close (Rebalance Trim)
On every rebalance cycle, positions are checked against the current ranking. If a held symbol has dropped out of the qualified rank window, its position is trimmed or closed. This is the primary momentum-fade exit — as soon as a coin stops being a relative outperformer, we exit.
Loss Cooldown
After a stop-loss fires on a symbol, a cooldown window prevents immediate re-entry into that symbol. The length varies per bot (8h for Bot8, 12h for Bot6). Re-entries after a stop-loss hit are statistically net-winners on the data, validating this as a conservative filter rather than a harmful blocker.
Manual emergency: A fleet-level emergency command (fleet_emergency.py --flatten --bot BotN) can instantly close all positions for a specific bot. This is the circuit breaker of last resort and is used only in systemic risk events, not for normal drawdowns.
This strategy is running live.
XS-Momentum bots have been active since May 2026. The Track Record page shows verified trade history, win rates, and drawdown data across all lookback variants.
Academic Reference
Han, B., Kang, J., & Ryu, D. (2024). Cross-sectional momentum in cryptocurrency markets. Validates persistent momentum effects in crypto perpetuals driven by retail flow, continuous price discovery, and funding rate feedback loops. Daniel, K. & Moskowitz, T.J. (2016). Momentum crashes. Journal of Financial Economics. Informs the BTC direction gate and crash-risk management framework used across all XS-MOM strategies.